Daniel Martí

I see. In that specific scenario I would say the YAML library is working as intended: it avoids producing output with trailing whitespace in lines, as that is very often...

Marcel found the same fix as @rudifa for a different evalv3 panic in https://review.gerrithub.io/c/cue-lang/cue/+/1202885 :) So this should be resolved in master shortly.

This appears to be fixed for both the old and new evaluators now; the testscript below passes: ``` env CUE_EXPERIMENT=evalv3=0 ! exec cue vet cmp stderr stderr.golden env CUE_EXPERIMENT=evalv3=1 !...

Also, the bisection above would make this issue a duplicate of https://github.com/cue-lang/cue/issues/3418, which makes complete sense. There, @rogpeppe said that a matchN error showed up on evalv3 but not on...

> Rather that _without_ `-c`, we still get errors when we should not, correct? Correct. > (notice I have hand-crafted the output, because testscript ordinarily fails on the first failure)...

It's perhaps also worth noting that the thread is a bit confusing now that we have two slightly different test scripts and manually crafted outputs :) It's fine by me...

> allows execution to continue in the presence of errors I guess its usefulness would be rather limited, because some errors really are fatal and will break the following commands...

The discussion about testscript was purely about providing a useful and self-describing reproduction script :) it has nothing to do with the fix itself.

@rogpeppe pointed out that the flag currently has three states: https://github.com/cue-lang/cue/blob/48207fb5fbfd85784c0560ea32c559183b49f475/cmd/cue/cmd/vet.go#L113-L118 That is, the presence of the flag changes behavior, it's not just its true/false value. So we then have...

My last comment here is correct - I still think the three modes are slightly confusing and could be better documented - but we should focus on the bug at...